Notes: Building Cursor At 23, Taking On GitHub Copilot

This post summarizes Michael Truell’s interview at Y Combinator (September 2025) about building Cursor, the AI-powered code editor that’s challenging GitHub Copilot.

The Origin Story

Michael Truell and his co-founders started Cursor when they were 23 years old. They didn’t begin with coding tools - their first project was actually about CAD (Computer-Aided Design).

The CAD Pivot

The team was initially excited about applying AI to CAD and 3D modeling. They spent time scraping CAD models from the internet, dealing with various file formats, and trying to convert them into something canonical. However, they realized they were really inherently excited about the future of coding.

They took a step back and realized that if they were being really consistent with their beliefs about where AI was heading, they should work on something they were genuinely passionate about - and that was coding.

Why Coding?

The team believed:

• AI would be transformative technology, perhaps more so than any technological revolution in recent centuries
• It would take a couple of decades to fully transform knowledge work
• There would be lots of incredibly valuable things to build over the next few decades
• Coding was the area they were most excited about

The VS Code Decision

One of the most critical decisions was whether to build a plugin for VS Code or fork it entirely.

Why Fork Instead of Plugin?

The team thought forking VS Code would be “a way, way, way longer road.” VS Code had been developed over 12 years, was one of the earliest TypeScript projects, and had lots of people working on it.

However, they realized that to build the best AI coding experience, they needed deep integration that a plugin couldn’t provide. The fork allowed them to:

• Control the entire user experience
• Integrate AI at every level of the editor
• Move faster on AI-specific features
• Not be constrained by plugin API limitations

The Competitive Landscape

When they started in 2022, working on AI was considered “kind of weird.” People weren’t entirely convinced that there would be lots of great applications from AI. Even people interested in AI were either:

• Focused on optimizing existing form factors (making products a little bit better)
• Asking “why would you work on anything other than AGI?”

The team took a middle road bet - believing there would be valuable products to build during the AI transition, not just at the end state.

Competing with GitHub Copilot

The interview touched on how Cursor differentiates from GitHub Copilot:

The Approach

• Deep integration: Rather than being a plugin, Cursor is a full editor built around AI
• User experience: Every interaction is designed with AI in mind
• Speed of iteration: As an independent company, they can move faster on AI features

The Market View

The team believes that for professional engineers:

• Code is still really important
• There will be a “long messy middle” where you work with AI more and more
• AI will become like a colleague
• It may also become like a very advanced compiler that can hide some of the code for you
• You’ll still need to read the logic, review it, and edit it

Building the Team

Staying Small

The team was very patient early on with hiring. They ended 2023 with only single-digit employees - less than 10 people. This was intentional:

• They believed they could go pretty far without hiring anyone
• They had their own set of missteps figuring out the first people to hire
• They focused on hiring a lot less than they probably should have early on

Hiring Philosophy

When hiring, they looked for people who:

• Were genuinely interested in the problem
• Could work independently
• Shared their vision for the future of coding

The Future of Coding

The Long View

Michael believes:

• AI is going to be transformative technology
• It will take a couple of decades to fully transform software development
• It’s an industry-wide effort with many independent capabilities that need to develop
• There are lots of incredibly valuable things to build

What Skills Still Matter?

When asked what skills will still matter:

“Programming, like math, is kind of just a good general education. I don’t think that goes away.”

He also noted:

• There are lots of practical skills from studying computer science
• The specific stuff studied in school isn’t super crucial in dynamic industries
• It’s more about the kind of learning you get along the way
• This hasn’t changed with AI

Advice for Engineering Students

For young people who want to follow a similar path:

Focus on What Interests You

“I think just working on things that you’re interested in and doing it with people both that you enjoy being around but that you respect a ton. And taking that really seriously.”

Avoid Box-Checking

For people in school, there are so many things that pull you toward:

• Checking boxes
• Less focusing on building something up over time

Instead, focus on:

• Building something you’re genuinely interested in
• Working with people you respect
• Taking your work seriously

Key Takeaways

1. Follow your genuine interests - The team pivoted from CAD to coding because that’s what they were truly excited about
2. Make bold technical decisions - Forking VS Code seemed like the longer road but enabled deeper AI integration
3. Stay small and focused - Less than 10 people through 2023 allowed them to move fast
4. Take the middle road - Not just optimizing existing products, not waiting for AGI, but building valuable tools during the transition
5. Programming education still matters - Like math, it’s good general education that won’t go away
6. Work with people you respect - Both enjoying their company and respecting their abilities